1: \begin{abstract}

2: Well-Tempered Metadynamics (WTmetaD) is an efficient method to enhance the reconstruction of the  

3: free-energy surface of proteins. WTmetaD guarantees a faster convergence in the long time

4:  limit in comparison with the standard metadynamics. It still suffers however from the same limitation, 

5: \textit{i.e.} the non trivial choice of pertinent collective variables (CVs). 

6: To circumvent this problem, we couple WTmetaD with a set of CVs generated from a dihedral Principal Component Analysis (dPCA) 

7: on the Ramachadran dihedral angles describing the backbone structure of the protein. The dPCA provides a generic method 

8: to extract relevant CVs built from internal coordinates. We illustrate the robustness of this method in the case of the small 

9: and very diffusive Met-enkephalin pentapeptide, and highlight a criterion to limit the number of CVs necessary to biased the 

10: metadynamics simulation. The free-energy landscape (FEL) of Met-enkephalin built on CVs generated from 

11: dPCA is found rugged compared with the FEL built on CVs extracted from PCA of the Cartesian coordinates of the atoms.

12: \end{abstract}

13: